This study was conducted in accordance with the Cochrane handbook for systematic reviews of diagnostic test accuracy, and has been reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA-DTA) statement, a copy of which is attached to this article as an Appendix (Appendix 1)19–21. The protocol for this review was prospectively registered with PROSPERO (ID CRD42021290610) and has been submitted for peer reviewed publication, with a pre-print available online22.
Search strategy and selection criteria
Studies meeting the following criteria were considered for inclusion:
- Participants: All post-operative patients over 18, of any operation type. No restrictions were placed on the study setting or length of follow-up.
- Index tests: Telemedicine by any method (telephone, photograph or questionnaire), including the use of questionnaires as these can be delivered remotely.
- Reference Standards: Face to face review, as per the United States (US) Centres for Disease Control and Prevention (CDC) criteria for SSI is deemed the gold standard, but no restrictions were placed if other methods were use. This was to ensure all available evidence would be synthesised.
- Target condition: SSI as defined by the CDC criteria; infection within 30 days of surgery or within 90 days if an implant is left in place23.
- Study design: Abstracts, reviews and conference proceeding were excluded. All other research designs were included in the systematic review, but only comparative, paired methodologies were taken forward to meta-analysis as all patients would experience index tests and reference standards.
Studies were excluded if they did not meet the inclusion criteria or were not presented in English (for lack of resources to translate other languages). The following databases were searched from inception to January 2022: Medline, Embase, CENTRAL and CINAHL. A combination of synonyms related to the keywords; “telemedicine” AND “surgical wound infection” formulated the terms used. The strategy used for Medline, Embase and CINAHL can be found in appendix 2.
The search strategy was developed with and conducted by an information specialist who uploaded results onto the Rayyan, a bespoke tool for conducting systematic reviews24. These were deduplicated before screening of titles and abstracts by two independent reviewers against the inclusion criteria. Relevant manuscripts were retrieved for full text review, and assessed for eligibility by two independent reviewers. Reference lists of these articles were searched manually for any additional studies not identified in preliminary search. Any disagreement at each stage was resolved by a third reviewer for consensus.
There were no limitations placed on study design for qualitative synthesis to comprehensively synthesise the literature. Reports with paired designs were taken forward for quantitative analysis to enable random-effects bivariate meta-analysis, and summary receiver operator characteristic (SROC) curves to be plotted.
Data extraction
A bespoke data spreadsheet (Microsoft Excel Version 16.59) was designed and utilised for data extraction by two independent authors. Data on study and diagnostic characteristics (author, year, country, study design, sample size, gender, age, telemedicine method, reference standard, type of surgery, follow-up schedule) among potential confounding factors (diabetes, BMI, and smoking status) were collected in addition to SSI rates, sensitivity, and specificity of diagnosis.
Surgical site infections were defined as per CDC criteria2. Only superficial SSI were included due to inherent barriers of diagnosing deep SSI remotely. No restrictions were placed on classification of telemedicine, reference standard type, or other characteristics.
Assessment of methodological quality
Risk of bias and the applicability of studies were assessed again by two independent reviewers with the QUADAS-2 tool25. Agreement of 80% across all categories on two included studies was considered sufficient before further assessment of remaining studies, as recommended by the Cochrane handbook for systematic reviews of diagnostic test accuracy20. Risk of bias and applicability scores were taken into consideration for subgroup meta-analysis, ascertaining a strength of recommendation from data retrieved.
Statistical analysis
Continuous descriptive characteristics were expressed as weighted mean averages with standard error. A bivariate model for meta-analysis was used to produce summary measures of sensitivity and specificity with confidence regions. All studies with paired designs had pooled forest plots and summary receiver operator characteristic curves synthesised in the initial exploratory analysis. Analysis was conducted with MetaDTA and plots constructed with Review Manager 5.426 27. Additional sources of heterogeneity were investigated through covariates (study country, type of surgery, telemedicine method, reference standard used).
For cases of multi-threshold test positivity, the cut-off achieving the maximum possible sensitivity – specificity trade off were taken forward. Indeterminate index test results were classified as ‘no SSI’ as this more closely reflects what would happen in practice. Tests were grouped as a unified ‘telemedicine’ and through the sub-groups; ‘photograph,’ ‘telephone,’ and ‘questionnaire.’ No studies reported video-based methods.
Subgroup analysis
All studies which compared photograph to face to face review will be referred to as photograph based telemedicine methods. Photograph-based methods utilise visual input whereas questionnaire and telephone do not incorporate trained physicians viewing a patient’s wound. As such, pre-specified analysis is conducted for studies including these methods for their sensitivity and specificity. Further analyses are performed as per the reference standard used and whether a pre-specified threshold was stated.